Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Limited prunable parameters (like shortcut, ..) #1

Open
kiddj opened this issue May 14, 2020 · 1 comment
Open

Limited prunable parameters (like shortcut, ..) #1

kiddj opened this issue May 14, 2020 · 1 comment

Comments

@kiddj
Copy link

kiddj commented May 14, 2020

Uploaded codes in repo does not seemed to support some parameters such as shortcut, batch-norm and bias terms.

Does LAP only work on weights except above terms?

@jaeho-lee
Copy link

Hi-

thank you for your interest in our work!

Our current repo only supports pruning weights (which dominate the number of parameters), but our framework can be smoothly extended to pruning shortcuts and batch-norms. The framework neglects bias terms.

We plan to make some updates -- after NeurIPS deadline -- to our codes to reflect the updates in the PyTorch 1.5.

Notes about pruning bias: There are some reasons to believe that either removing them all at once, or keeping them altogether won't be too detrimental to the performance. I like the discussion here: https://www.reddit.com/r/MachineLearning/comments/eymex9/d_why_isnt_the_bias_terms_in_the_weights_also/

We'll keep this issue opened until we update our package.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants